Using an LSTM and Classification Methods to Determine Risk of dB/dt Threshold Crossings as Proxy for Geomagnetically Induced Currents

Conference Paper


  • The interaction between the solar wind and the Magnetosphere can produce Geomagnetically Induced Currents (GIC’s) on the ground, which are capable of causing power outages and damage to crucial infrastructure. • The ability to predict when and where these events may occur could allow us to avoid the worst of this damage. • The use of physics-informed machine learning models can offer a computationally inexpensive method of predicting GIC events using horizontal dB/dt as a proxy, though most models thus far have fallen short of consistently accurate predictions. dB/dt was defined as: dB_H/dt = sqrt(dE^2/dt + dN^2/dt) With N and E the North and East components of the magnetic field respectively. • Here, a Long-Short Term Memory (LSTM) model was used to determine the risk of dB/dt going over thresholds of 9, 18, 42, 66, and 90 nT/min for the Ottawa (OTT) ground magnetometer station. • Three storms were chosen for testing and removed from the training set: March 30, 2001 (~ -211nT), December 14, 2006 (~ -437nT), & August 05, 2011 (~ -126nT). • The storms were chosen for several reasons; they represent different storm intensities, they occurred at different points in the solar cycle, and there are minimal gaps in the data that needed to be interpolated over.
  • Authors

  • Johnson, Jeremiah
  • Coughlan, Michael
  • Keesee, Amy
  • Pinto, Victor
  • Connor, Hyunju